1. Field of the Invention
The present invention relates to an image processing device that performs image processing on an image obtained by using an X-ray CT apparatus and so forth, and more particularly relates to a technology that maintains an original shape of the edge of a structure and removes streak artifacts.
2. Description of the Related Art
In an image reconstructing method, the X-ray CT apparatus reconstructs a tomographic image of a test object by irradiating the test object with X-rays from around the object and back-projecting real projection data acquired at a plurality of projection angles.
When an image of flat body part such as the shoulder and so forth of the test object is to be reconstructed on the basis of original information obtained by such an X-ray CT apparatus as aforementioned, in a case where an exposure dose of the X-rays is insufficient, the streak artifacts are generated in a direction which is large in X-ray attenuation. A technique of reducing the streak artifacts by smoothing projection data and thereafter applying the aforementioned image reconstructing method is proposed in J. Hsieh, “Adaptive streak artifact reduction in computed tomography resulting from excessive x-ray photon noise”, Med. Phys. Vol. 25, No. 11, pp. 2139-2147, 1998 and T. Li et. al., “Nonlinear Sinogram Smoothing for Low-Dose X-ray CT”, IEEE. Trans. Nucl. Sci., Vol. 51, No. 5, pp. 2505-2513, 2004. Hereinafter, a device for reducing these streak artifacts will be referred to as a “projection data smoothing filter”.
In an image obtained by the projection data smoothing filter, an edge of a comparatively high-contrast structure is blurred in association with attainment of a streak artifact reducing effect. The streak artifact reducing effect and a structure edge maintaining effect are in a trade-off relation and the more the former effect is increased, the more the latter effect is decreased. Accordingly, in a case where strong streak artifacts are reduced, the edge of the structure may be greatly blurred. Since the blur on the edge of the structure is not favorable in clinical diagnosis, an image (hereinafter, referred to as a “target image”) that the edge of the structure is maintained and the streak artifacts are removed is expected.
As a means of settling the above-mentioned disadvantage, a method of appropriately adding together with weighting an image (hereinafter, referred to as an “original image”) obtained from the projection data by the aforementioned image reconstructing method and an image (hereinafter, referred to as a “smoothed image”) obtained from projection data which has been smoothed by application of the projection data smoothing filter by the aforementioned image reconstructing method is given. That is, the aforementioned target image is obtained by performing weighted addition such that the percentage of a pixel value of the smoothed image is increased in pixels of a part that the streaked artifact is generated on the image and the percentage of a pixel value of the original image is increased in pixels of the edge part of the structure. Several techniques of mixing together two images which are generated from the same original information as mentioned above have been proposed so far.
In the technique described in Japanese Patent No. 3700798, a low-resolution image which is lower in resolution than the original image is created from the original image, the edge is detected from the low-resolution image and a low-resolution edge image is created, and a weighting coefficient is calculated from an edge luminance value of the low-resolution edge image using a nonlinear function. At that time, a difference between a maximum value and a minimum value of values of pixels in a predetermined region including a pixel of interest, and a variance value, a concentration gradient and so forth thereof are used in detection of the edge. The edge of the structure is enhanced by adding an image that the low-resolution image has been differentiated from the original image to the original image with weighting by using the aforementioned weighting coefficient. It is possible to perform weighted addition of the original image and the smoothed image by such a technique.
In the technique described in Japanese Patent Application Laid-Open No. Hei07-93543, an image is divided into small regions, a standard deviation value of the small regions is calculated, a histogram that the standard deviation value is used for classification targeting on all of the small regions is prepared, and a discrimination class value is calculated from the histogram to classify each small region into a flat region and a non-flat region. Detecting the edge of the structure on the image is equivalent to discriminating between the flat region and the non-flat region. If it is possible to classify each small region into both of the flat region and the non-flat region, it will become possible to perform weighted addition of the original image and the smoothed image by using the pixel value of the original image in the non-flat region and using the pixel value of the smoothed image in the flat region.